Words not in CountVectorizer vocab despite being well above min_df threshold

Nobody has claimed this yet.

Hi Maarten,

I'm having an issue with some important words not appearing within CountVectorizer when using `min_df` even though they are well above the set threshold. My understanding of `min_df` is that when it is an integer (say, 10), all words that appear in at least 10 documents will be kept. Is that correct? See the example below,

```python
from umap import UMAP
from datasets import load_dataset
from bertopic import BERTopic
from sklearn.feature_extraction.text import CountVectorizer

dataset = load_dataset("CShorten/ML-ArXiv-Papers")["train"]
abstracts = dataset["abstract"][:5_000]

umap_model = UMAP(n_neighbors=15, n_components=5, min_dist=0.0, metric='cosine', random_state=42)
vectorizer_model=CountVectorizer(min_df=10)

topic_model = BERTopic(umap_model=umap_model, vectorizer_model=vectorizer_model).fit(abstracts)
```

Take the word 'cancer' for example.

```python
# this under counts as it doesn't consider words next to punctuation,
# but wont consider words inside words (e.g. robot in robotics)
len([d for d in abstracts if " cancer " in d.lower()])
>>> 32
```

'cancer' appears in at least 32 documents (not counting occurrences up against punctuation so should be well above `min_df`.

```python
topic_model.vectorizer_model.vocabulary_["cancer"]
```
returns a key error.

Also try words: *privacy, reward, quantum, fuzzy, hashing, planning, player, robot, embeddings*

If I remove the min_df param and just use `CountVectorizer()`, then these words are all present.

I first came across this with the word 'rust' which was an important word in a large topic with ~70 document occurrences and `min_df=10`. It isn't prominent in the arxiv dataset though.

Do you know what is happening here?

Start with the provided 5,000-abstract reproduction using BERTopic.fit, UMAP, and CountVectorizer(min_df=10). Compare the resulting vocabulary with a plain CountVectorizer and inspect how BERTopic processes the vectorizer output. Done means explaining the missing terms and confirming corrected min_df behavior without breaking the reported example.

Written by the indexing model from the issue text.